Generative Adversarial Network Sculptures
A Generative Adversarial Network (GAN) is a system which generates new data through dynamic learning. The GAN creates images (here, generated from a pool of images of my past work). These images are then run through a discriminative network, which deciphers whether the created image is similar enough to the source. It works in this manner to continually produce more convincing images. Working with GAN, I too am a discriminator in this process, I decide.
“Yes, I like this image.”
“No, it doesn’t look like anything I’ve made before.”
“Yes, I will take that image and turn it into something that I would like to make.”
And like that, a computer-human relationship and collaboration is formed. Here, I work with 2-dimensional digital files created by the GAN and create my own version of the GAN’s output. I have only one side of the form to work from, so I must interpret the unknown intuitively.
“Yes, I like this image.”
“No, it doesn’t look like anything I’ve made before.”
“Yes, I will take that image and turn it into something that I would like to make.”
And like that, a computer-human relationship and collaboration is formed. Here, I work with 2-dimensional digital files created by the GAN and create my own version of the GAN’s output. I have only one side of the form to work from, so I must interpret the unknown intuitively.
